Developers Can Now Try Out the AI API Gateway

Since Stanford’s AI Playground launched in fall 2024, many early adopters have asked about direct API (application programming interface) access, in hopes of developing custom tools and solutions. 

Now, we are excited to announce the launch of the new AI API Gateway service—designed for developers and tinkerers who want to create custom tools and integrations using the models available through the AI Playground.

About the AI Playground

Not sure what the AI Playground is exactly? In short, it’s a Stanford-managed platform built on open-source technology that provides access to various large language models (LLMs). The service is available to Stanford faculty, students, and staff, providing a more secure way to learn and experiment with popular AI models and tools.

Developer notes

With the AI API Gateway service, you can take the potential of our available AI models and tools even further. A few opportunities that have come up are:

• Custom chatbots and virtual assistants: Build chatbots or virtual assistants tailored to specific tasks like answering FAQs or looking up data on your behalf.
• Intelligent data analytics tools: Integrate AI-driven text summarization and analysis into platforms to efficiently handle large data collections.
• Enhanced customer support: Embed AI-based responses into support systems or websites to answer common user questions automatically.
• Automated content creation: Create custom AI solutions to help generate documents for existing systems and webpages.

If you plan to work with this API service, please remember to always keep a human in the loop when designing any automation with AI. Human oversight is the first of the seven Guiding Principles for AI at Stanford.

Also, please remember: the AI Playground—and consequently the new AI API Gateway service—is not currently approved for High Risk Data, personally identifiable information (PII), or protected health information (PHI).

Getting started with your API key

Ready to dive in? We suggest to start with cost planning. You'll be asked to indicate a max budget, which prevents API calls beyond that budget. Be sure to:

• Review the AI API Gateway Rates.
• Have an approved PTA for this purpose.

Next, follow these steps for the technical implementation:

1. Request the creation of a new API key.
2. Once approved by the PTA owner, your API key will be securely sent to you. (This setup process could take two to five days.)
3. Follow the guidance in the AI Playground Access and Usage help page.

You can find more information in the AI API Gateway FAQs or by sending a help request.

Moving forward together

We can’t wait to see what our Stanford community is able to discover with this new way to work with AI. Share your progress, ideas, and questions for others using the AI Playground in the #ai-playground Slack channel.